MLBEDSW-10828: Align int16 quantized scale to reference (!283) · Merge requests · artificial-intelligence / ethos-u / Vela

Max Bergfelt requested to merge MLBEDSW-10828 into main Jul 15, 2025

Changed the calculation of the scale in the QuantizedScale function to bit-exactly match the reference, in which the scale is calculated in two steps for int16 instead of one.

Change-Id: I1f72a58b3b3b7abb93ff1a4096e7be9897bd2879 Signed-off-by: Max Bergfelt max.bergfelt@arm.com

MLBEDSW-10828: Align int16 quantized scale to reference

Merge request reports